If the cluster becomes unavailable, ceph_osdc_notify() may hang even
with osd_request_timeout option set because linger_notify_finish_wait()
waits for MWatchNotify NOTIFY_COMPLETE message with no associated OSD
request in flight -- it's completely asynchronous.
Introduce an additional timeout, derived from the specified notify
timeout. While at it, switch both waits to killable which is more
correct.
Cc: stable@vger.kernel.org
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Dongsheng Yang <dongsheng.yang@easystack.cn>
Reviewed-by: Xiubo Li <xiubli@redhat.com>
Due to rbd_try_acquire_lock() effectively swallowing all but
EBLOCKLISTED error from rbd_try_lock() ("request lock anyway") and
rbd_request_lock() returning ETIMEDOUT error not only for an actual
notify timeout but also when the lock owner doesn't respond, a busy
loop inside of rbd_acquire_lock() between rbd_try_acquire_lock() and
rbd_request_lock() is possible.
Requesting the lock on EBUSY error (returned by get_lock_owner_info()
if an incompatible lock or invalid lock owner is detected) makes very
little sense. The same goes for ETIMEDOUT error (might pop up pretty
much anywhere if osd_request_timeout option is set) and many others.
Just fail I/O requests on rbd_dev->acquiring_list immediately on any
error from rbd_try_lock().
Cc: stable@vger.kernel.org # 588159009d: rbd: retrieve and check lock owner twice before blocklisting
Cc: stable@vger.kernel.org
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Dongsheng Yang <dongsheng.yang@easystack.cn>
The one function left in this file does not justify a whole extra
file nor the keystone.h support file used to forward declare that
function. Merge this into the main keystone.c file.
Signed-off-by: Andrew Davis <afd@ti.com>
Link: https://lore.kernel.org/r/20230726160441.101566-2-afd@ti.com
Signed-off-by: Nishanth Menon <nm@ti.com>
The defines in memory.h can go into the file that makes use of them.
No reason to have a header here, remove.
Signed-off-by: Andrew Davis <afd@ti.com>
Link: https://lore.kernel.org/r/20230726160441.101566-1-afd@ti.com
Signed-off-by: Nishanth Menon <nm@ti.com>
Drop the non-fatal probe deferral log for getting MSI domain.
This makes the kernel log clean and we do not get recurring logs
stating: "Failed to get MSI domain".
Signed-off-by: Jayesh Choudhary <j-choudhary@ti.com>
Link: https://lore.kernel.org/r/20230728053356.31044-1-j-choudhary@ti.com
Signed-off-by: Nishanth Menon <nm@ti.com>
Enable TI ECAP, DP83869 driver, OMAP2 Mailbox, K3 remote proc
SND_SOC_J721E_EVM, MCAN, UFS and RTI, Thermal driver to be built
as module.
These configs are supported on below TI platforms
ECAP on am642, am65 iot2050 EVM.
DP83869 on AM64x EVM.
OMAP2 Mailbox, K3 remote proc on AM64 on J7200,J721E and J784S4 SOC.
CONFIG_SND_SOC_J721E_EVM on J721E.
MCAN on AM68 SK, J721E, J721S2, AM642 AM642-Phyboards
UFS feature on J721E EVM
RTI watchdog on AM62, AM62A, AM64, AM65, J721E and J7200 SOC.
Thermal driver on AM62, AM62A AM65, J7200, J721E, J721S2,
J784S4 platform.
Signed-off-by: Udit Kumar <u-kumar1@ti.com>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/r/20230731161048.2522154-1-u-kumar1@ti.com
Signed-off-by: Nishanth Menon <nm@ti.com>
The MAIN CPSW2G instance of CPSW on J721S2 SoC can be enabled with the GESI
Expansion Board connected to the J7 Common-Proc-Board. Use the overlay
to enable this.
Add alias for the MAIN CPSW2G port to enable kernel to fetch MAC address
directly from U-Boot.
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
Signed-off-by: Siddharth Vadapalli <s-vadapalli@ti.com>
Reviewed-by: Ravi Gunasekaran <r-gunasekaran@ti.com>
Link: https://lore.kernel.org/r/20230726065407.378455-3-s-vadapalli@ti.com
Signed-off-by: Nishanth Menon <nm@ti.com>
TI's J721S2 SoC has a MAIN CPSW2G instance of the CPSW Ethernet Switch.
Add devicetree node for it.
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>
Signed-off-by: Siddharth Vadapalli <s-vadapalli@ti.com>
Reviewed-by: Ravi Gunasekaran <r-gunasekaran@ti.com>
Link: https://lore.kernel.org/r/20230726065407.378455-2-s-vadapalli@ti.com
Signed-off-by: Nishanth Menon <nm@ti.com>
The J7 GESI EXP board for J721E Common-Proc-Board supports RGMII mode.
Use the overlay to configure CPSW9G ports in RGMII-RXID mode.
Add aliases for CPSW9G ports to enable kernel to fetch MAC addresses
directly from U-Boot.
Signed-off-by: Siddharth Vadapalli <s-vadapalli@ti.com>
Reviewed-by: Ravi Gunasekaran <r-gunasekaran@ti.com>
Link: https://lore.kernel.org/r/20230725073057.96705-1-s-vadapalli@ti.com
Signed-off-by: Nishanth Menon <nm@ti.com>
Add dts nodes for 6 EHRPWM instances on SoC. Disable EHRPWM nodes in the
dtsi files and only enable the ones that are actually pinned out on a
given board in the board dts file.
Signed-off-by: Sinthu Raja <sinthu.raja@ti.com>
Link: https://lore.kernel.org/r/20230721082150.12599-1-sinthu.raja@ti.com
Signed-off-by: Nishanth Menon <nm@ti.com>
CAN instances 3 and 5 in the main domain are brought on the common
processor board through header J27 and J28. The CAN High and Low lines
from the SoC are routed through a mux on the SoM. The select lines need
to be set for the CAN signals to get connected to the transceivers on
the common processor board. Threfore, add respective mux, transceiver
dt nodes to add support for these CAN instances.
Reviewed-by: Udit Kumar <u-kumar1@ti.com>
Signed-off-by: Bhavya Kapoor <b-kapoor@ti.com>
Link: https://lore.kernel.org/r/20230725085939.536766-1-b-kapoor@ti.com
Signed-off-by: Nishanth Menon <nm@ti.com>
Introduce the debounce select mux macros to allow folks to setup
debounce configuration for pins. Each configuration selected maps
to a specific timing register as documented in appropriate Technical
Reference Manual (example:[1]).
[1] AM625x TRM (section 6.1.2.2): https://www.ti.com/lit/pdf/spruiv7
Reviewed-by: Vignesh Raghavendra <vigneshr@ti.com>
Link: https://lore.kernel.org/r/20230619131620.3286650-1-nm@ti.com
Signed-off-by: Nishanth Menon <nm@ti.com>
The dcbnl_bcn_setcfg uses erroneous policy to parse tb[DCB_ATTR_BCN],
which is introduced in commit 859ee3c438 ("DCB: Add support for DCB
BCN"). Please see the comment in below code
static int dcbnl_bcn_setcfg(...)
{
...
ret = nla_parse_nested_deprecated(..., dcbnl_pfc_up_nest, .. )
// !!! dcbnl_pfc_up_nest for attributes
// DCB_PFC_UP_ATTR_0 to DCB_PFC_UP_ATTR_ALL in enum dcbnl_pfc_up_attrs
...
for (i = DCB_BCN_ATTR_RP_0; i <= DCB_BCN_ATTR_RP_7; i++) {
// !!! DCB_BCN_ATTR_RP_0 to DCB_BCN_ATTR_RP_7 in enum dcbnl_bcn_attrs
...
value_byte = nla_get_u8(data[i]);
...
}
...
for (i = DCB_BCN_ATTR_BCNA_0; i <= DCB_BCN_ATTR_RI; i++) {
// !!! DCB_BCN_ATTR_BCNA_0 to DCB_BCN_ATTR_RI in enum dcbnl_bcn_attrs
...
value_int = nla_get_u32(data[i]);
...
}
...
}
That is, the nla_parse_nested_deprecated uses dcbnl_pfc_up_nest
attributes to parse nlattr defined in dcbnl_pfc_up_attrs. But the
following access code fetch each nlattr as dcbnl_bcn_attrs attributes.
By looking up the associated nla_policy for dcbnl_bcn_attrs. We can find
the beginning part of these two policies are "same".
static const struct nla_policy dcbnl_pfc_up_nest[...] = {
[DCB_PFC_UP_ATTR_0] = {.type = NLA_U8},
[DCB_PFC_UP_ATTR_1] = {.type = NLA_U8},
[DCB_PFC_UP_ATTR_2] = {.type = NLA_U8},
[DCB_PFC_UP_ATTR_3] = {.type = NLA_U8},
[DCB_PFC_UP_ATTR_4] = {.type = NLA_U8},
[DCB_PFC_UP_ATTR_5] = {.type = NLA_U8},
[DCB_PFC_UP_ATTR_6] = {.type = NLA_U8},
[DCB_PFC_UP_ATTR_7] = {.type = NLA_U8},
[DCB_PFC_UP_ATTR_ALL] = {.type = NLA_FLAG},
};
static const struct nla_policy dcbnl_bcn_nest[...] = {
[DCB_BCN_ATTR_RP_0] = {.type = NLA_U8},
[DCB_BCN_ATTR_RP_1] = {.type = NLA_U8},
[DCB_BCN_ATTR_RP_2] = {.type = NLA_U8},
[DCB_BCN_ATTR_RP_3] = {.type = NLA_U8},
[DCB_BCN_ATTR_RP_4] = {.type = NLA_U8},
[DCB_BCN_ATTR_RP_5] = {.type = NLA_U8},
[DCB_BCN_ATTR_RP_6] = {.type = NLA_U8},
[DCB_BCN_ATTR_RP_7] = {.type = NLA_U8},
[DCB_BCN_ATTR_RP_ALL] = {.type = NLA_FLAG},
// from here is somewhat different
[DCB_BCN_ATTR_BCNA_0] = {.type = NLA_U32},
...
[DCB_BCN_ATTR_ALL] = {.type = NLA_FLAG},
};
Therefore, the current code is buggy and this
nla_parse_nested_deprecated could overflow the dcbnl_pfc_up_nest and use
the adjacent nla_policy to parse attributes from DCB_BCN_ATTR_BCNA_0.
Hence use the correct policy dcbnl_bcn_nest to parse the nested
tb[DCB_ATTR_BCN] TLV.
Fixes: 859ee3c438 ("DCB: Add support for DCB BCN")
Signed-off-by: Lin Ma <linma@zju.edu.cn>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20230801013248.87240-1-linma@zju.edu.cn
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Radu Pirea says:
====================
Add TJA1120 support
This patch series got bigger than I expected. It cleans up the
next-c45-tja11xx driver and adds support for the TJA1120(1000BaseT1
automotive phy).
Master/slave custom implementation was replaced with the generic
implementation (genphy_c45_config_aneg/genphy_c45_read_status).
The TJA1120 and TJA1103 are a bit different when it comes to the PTP
interface. The timestamp read procedure was changed, some addresses were
changed and some bits were moved from one register to another. Adding
TJA1120 support was tricky, and I tried not to duplicate the code. If
something looks too hacky to you, I am open to suggestions.
====================
Link: https://lore.kernel.org/r/20230731091619.77961-1-radu-nicolae.pirea@oss.nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
During PTP testing on early TJA1120 engineering samples I observed that
if the link is lost and recovered, the tx timestamps will be randomly
lost. To avoid this HW issue, the PCS should be reset.
Resetting the PCS will break the link and we should reset the PCS on
LINK UP -> LINK DOWN transition, otherwise we will trigger and infinite
loop of LINK UP -> LINK DOWN events.
Signed-off-by: Radu Pirea (NXP OSS) <radu-nicolae.pirea@oss.nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20230731091619.77961-12-radu-nicolae.pirea@oss.nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
On TJA1120, the external trigger timestamp now has a VALID bit. This
changes the logic and we can't use the TJA1103 procedure.
For TJA1103, we can always read a valid timestamp from the registers,
compare the new timestamp with the old timestamp and, if they are not the
same, an event occurred. This logic cannot be applied for TJA1120 because
the timestamp is 0 if the VALID bit is not set.
Signed-off-by: Radu Pirea (NXP OSS) <radu-nicolae.pirea@oss.nxp.com>
Link: https://lore.kernel.org/r/20230731091619.77961-11-radu-nicolae.pirea@oss.nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
For TJA1120, the enable bit for cable test is not writable if the PHY is
not in test mode.
Signed-off-by: Radu Pirea (NXP OSS) <radu-nicolae.pirea@oss.nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20230731091619.77961-10-radu-nicolae.pirea@oss.nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
TJA1120 and TJA1103 have a set of functional safety hardware tests
executed after every reset, and when the tests are done, the IRQ line is
asserted. For the moment, the purpose of these handlers is to acknowledge
the IRQ and not to check the FUSA tests status.
Signed-off-by: Radu Pirea (NXP OSS) <radu-nicolae.pirea@oss.nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20230731091619.77961-9-radu-nicolae.pirea@oss.nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The egress timestamp FIFO/circular buffer work different on TJA1120 than
TJA1103.
For TJA1103 the new timestamp should be manually moved from the FIFO to
the hardware buffer before checking if the timestamp is valid.
For TJA1120 the hardware will move automatically the new timestamp
from the FIFO to the buffer and the user should check the valid bit, read
the timestamp and unlock the buffer by writing any of the buffer
registers(which are read only).
Another change for the TJA1120 is the behaviour of the EGR TS IRQ bit.
This bit was a self-clear bit for TJA1103, but now should be cleared
before reading the timestamp.
Signed-off-by: Radu Pirea (NXP OSS) <radu-nicolae.pirea@oss.nxp.com>
Link: https://lore.kernel.org/r/20230731091619.77961-8-radu-nicolae.pirea@oss.nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The external trigger configuration for TJA1120 has changed. The PHY
supports sampling of the LTC on rising and on falling edge.
Signed-off-by: Radu Pirea (NXP OSS) <radu-nicolae.pirea@oss.nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20230731091619.77961-7-radu-nicolae.pirea@oss.nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
PHY_BASIC_T1_FEATURES are not the right features supported by TJA1103
anymore.
For example ethtool reports:
[root@alarm ~]# ethtool end0
Settings for end0:
Supported ports: [ TP ]
Supported link modes: 100baseT1/Full
10baseT1L/Full
10baseT1L/Full is not supported by TJA1103 and supported ports list is
not completed. The PHY also have a MII port.
genphy_c45_pma_read_abilities implementation can detect the PHY features
and they look like this.
[root@alarm ~]# ethtool end0
Settings for end0:
Supported ports: [ TP MII ]
Supported link modes: 100baseT1/Full
Supported pause frame use: Symmetric
Supports auto-negotiation: No
Supported FEC modes: Not reported
Advertised link modes: 100baseT1/Full
Advertised pause frame use: Symmetric
Advertised auto-negotiation: No
Advertised FEC modes: Not reported
Speed: 100Mb/s
Duplex: Full
Auto-negotiation: off
master-slave cfg: forced master
master-slave status: master
Port: Twisted Pair
PHYAD: 1
Transceiver: external
MDI-X: Unknown
Supports Wake-on: g
Wake-on: d
Link detected: yes
SQI: 7/7
Signed-off-by: Radu Pirea (NXP OSS) <radu-nicolae.pirea@oss.nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20230731091619.77961-5-radu-nicolae.pirea@oss.nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Between TJA1120 and TJA1103 the hardware was improved, but some register
addresses were changed and some bit fields were moved from one register
to another.
Introduce the nxp_c45_reg_field structure and its associated functions to
abstract the differences between the PHYs.
Remove the defined bits and register addresses that are not common
between TJA1103 and TJA1120 and replace them with reg_fields and
register addresses from phydev->drv->driver_data.
Signed-off-by: Radu Pirea (NXP OSS) <radu-nicolae.pirea@oss.nxp.com>
Link: https://lore.kernel.org/r/20230731091619.77961-4-radu-nicolae.pirea@oss.nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Remove RX BIST frame counters from the PHY statistics.
In production mode, these counters are always read as 0.
Signed-off-by: Radu Pirea (NXP OSS) <radu-nicolae.pirea@oss.nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20230731091619.77961-3-radu-nicolae.pirea@oss.nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Remove the custom implementation of master/save setup and read status
and use genphy_c45_config_aneg and genphy_c45_read_status since phylib
has support for master/slave setup and master/slave status.
Signed-off-by: Radu Pirea (NXP OSS) <radu-nicolae.pirea@oss.nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20230731091619.77961-2-radu-nicolae.pirea@oss.nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Gavin Li says:
====================
virtio_net: add per queue interrupt coalescing support
Currently, coalescing parameters are grouped for all transmit and receive
virtqueues. This patch series add support to set or get the parameters for
a specified virtqueue.
When the traffic between virtqueues is unbalanced, for example, one virtqueue
is busy and another virtqueue is idle, then it will be very useful to
control coalescing parameters at the virtqueue granularity.
Example command:
$ ethtool -Q eth5 queue_mask 0x1 --coalesce tx-packets 10
Would set max_packets=10 to VQ 1.
$ ethtool -Q eth5 queue_mask 0x1 --coalesce rx-packets 10
Would set max_packets=10 to VQ 0.
$ ethtool -Q eth5 queue_mask 0x1 --show-coalesce
Queue: 0
Adaptive RX: off TX: off
stats-block-usecs: 0
sample-interval: 0
pkt-rate-low: 0
pkt-rate-high: 0
rx-usecs: 222
rx-frames: 0
rx-usecs-irq: 0
rx-frames-irq: 256
tx-usecs: 222
tx-frames: 0
tx-usecs-irq: 0
tx-frames-irq: 256
rx-usecs-low: 0
rx-frame-low: 0
tx-usecs-low: 0
tx-frame-low: 0
rx-usecs-high: 0
rx-frame-high: 0
tx-usecs-high: 0
tx-frame-high: 0
====================
Link: https://lore.kernel.org/r/20230731070656.96411-1-gavinl@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Enable per queue interrupt coalesce feature bit in driver and validate its
dependency with control queue.
Signed-off-by: Gavin Li <gavinl@nvidia.com>
Reviewed-by: Dragos Tatulea <dtatulea@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Heng Qi <hengqi@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Link: https://lore.kernel.org/r/20230731070656.96411-4-gavinl@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add interrupt_coalesce config in send_queue and receive_queue to cache user
config.
Send per virtqueue interrupt moderation config to underlying device in
order to have more efficient interrupt moderation and cpu utilization of
guest VM.
Additionally, address all the VQs when updating the global configuration,
as now the individual VQs configuration can diverge from the global
configuration.
Signed-off-by: Gavin Li <gavinl@nvidia.com>
Reviewed-by: Dragos Tatulea <dtatulea@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Heng Qi <hengqi@linux.alibaba.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Link: https://lore.kernel.org/r/20230731070656.96411-3-gavinl@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Extract interrupt coalescing settings to a structure so that it could be
reused in other data structures.
Signed-off-by: Gavin Li <gavinl@nvidia.com>
Reviewed-by: Dragos Tatulea <dtatulea@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Heng Qi <hengqi@linux.alibaba.com>
Link: https://lore.kernel.org/r/20230731070656.96411-2-gavinl@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
There is no possible for platform_get_irq() to return 0,
and the return value of platform_get_irq() is more sensible
to show the error reason.
Signed-off-by: Ruan Jinjie <ruanjinjie@huawei.com>
Link: https://lore.kernel.org/r/20230731122246.2028651-1-ruanjinjie@huawei.com
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
iomap always punts async dio write completions to a workqueue, which has
a cost in terms of efficiency (now you need an unrelated worker to
process it) and latency (now you're bouncing a completion through an
async worker, which is a classic slowdown scenario).
io_uring handles IRQ completions via task_work, and for writes that
don't need to do extra IO at completion time, we can safely complete
them inline from that. This patchset adds IOCB_DIO_CALLER_COMP, which an
IO issuer can set to inform the completion side that any extra work that
needs doing for that completion can be punted to a safe task context.
The iomap dio completion will happen in hard/soft irq context, and we
need a saner context to process these completions. IOCB_DIO_CALLER_COMP
is added, which can be set in a struct kiocb->ki_flags by the issuer. If
the completion side of the iocb handling understands this flag, it can
choose to set a kiocb->dio_complete() handler and just call ki_complete
from IRQ context. The issuer must then ensure that this callback is
processed from a task. io_uring punts IRQ completions to task_work
already, so it's trivial wire it up to run more of the completion before
posting a CQE. This is good for up to a 37% improvement in
throughput/latency for low queue depth IO, patch 5 has the details.
If we need to do real work at completion time, iomap will clear the
IOMAP_DIO_CALLER_COMP flag.
This work came about when Andres tested low queue depth dio writes for
postgres and compared it to doing sync dio writes, showing that the
async processing slows us down a lot.
-----BEGIN PGP SIGNATURE-----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=EnSA
-----END PGP SIGNATURE-----
Merge tag 'xfs-async-dio.6-2023-08-01' of git://git.kernel.dk/linux into iomap-6.6-mergeA
Improve iomap/xfs async dio write performance
iomap always punts async dio write completions to a workqueue, which has
a cost in terms of efficiency (now you need an unrelated worker to
process it) and latency (now you're bouncing a completion through an
async worker, which is a classic slowdown scenario).
io_uring handles IRQ completions via task_work, and for writes that
don't need to do extra IO at completion time, we can safely complete
them inline from that. This patchset adds IOCB_DIO_CALLER_COMP, which an
IO issuer can set to inform the completion side that any extra work that
needs doing for that completion can be punted to a safe task context.
The iomap dio completion will happen in hard/soft irq context, and we
need a saner context to process these completions. IOCB_DIO_CALLER_COMP
is added, which can be set in a struct kiocb->ki_flags by the issuer. If
the completion side of the iocb handling understands this flag, it can
choose to set a kiocb->dio_complete() handler and just call ki_complete
from IRQ context. The issuer must then ensure that this callback is
processed from a task. io_uring punts IRQ completions to task_work
already, so it's trivial wire it up to run more of the completion before
posting a CQE. This is good for up to a 37% improvement in
throughput/latency for low queue depth IO, patch 5 has the details.
If we need to do real work at completion time, iomap will clear the
IOMAP_DIO_CALLER_COMP flag.
This work came about when Andres tested low queue depth dio writes for
postgres and compared it to doing sync dio writes, showing that the
async processing slows us down a lot.
* tag 'xfs-async-dio.6-2023-08-01' of git://git.kernel.dk/linux:
iomap: support IOCB_DIO_CALLER_COMP
io_uring/rw: add write support for IOCB_DIO_CALLER_COMP
fs: add IOCB flags related to passing back dio completions
iomap: add IOMAP_DIO_INLINE_COMP
iomap: only set iocb->private for polled bio
iomap: treat a write through cache the same as FUA
iomap: use an unsigned type for IOMAP_DIO_* defines
iomap: cleanup up iomap_dio_bio_end_io()
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
If IOCB_DIO_CALLER_COMP is set, utilize that to set kiocb->dio_complete
handler and data for that callback. Rather than punt the completion to a
workqueue, we pass back the handler and data to the issuer and will get
a callback from a safe task context.
Using the following fio job to randomly dio write 4k blocks at
queue depths of 1..16:
fio --name=dio-write --filename=/data1/file --time_based=1 \
--runtime=10 --bs=4096 --rw=randwrite --norandommap --buffered=0 \
--cpus_allowed=4 --ioengine=io_uring --iodepth=$depth
shows the following results before and after this patch:
Stock Patched Diff
=======================================
QD1 155K 162K + 4.5%
QD2 290K 313K + 7.9%
QD4 533K 597K +12.0%
QD8 604K 827K +36.9%
QD16 615K 845K +37.4%
which shows nice wins all around. If we factored in per-IOP efficiency,
the wins look even nicer. This becomes apparent as queue depth rises,
as the offloaded workqueue completions runs out of steam.
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If the filesystem dio handler understands IOCB_DIO_CALLER_COMP, we'll
get a kiocb->ki_complete() callback with kiocb->dio_complete set. In
that case, rather than complete the IO directly through task_work, queue
up an intermediate task_work handler that first processes this callback
and then immediately completes the request.
For XFS, this avoids a punt through a workqueue, which is a lot less
efficient and adds latency to lower queue depth (or sync) O_DIRECT
writes.
Only do this for non-polled IO, as polled IO doesn't need this kind
of deferral as it always completes within the task itself. This then
avoids a check for deferral in the polled IO completion handler.
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Async dio completions generally happen from hard/soft IRQ context, which
means that users like iomap may need to defer some of the completion
handling to a workqueue. This is less efficient than having the original
issuer handle it, like we do for sync IO, and it adds latency to the
completions.
Add IOCB_DIO_CALLER_COMP, which the issuer can set if it is able to
safely punt these completions to a safe context. If the dio handler is
aware of this flag, assign a callback handler in kiocb->dio_complete and
associated data io kiocb->private. The issuer will then call this
handler with that data from task context.
No functional changes in this patch.
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Rather than gate whether or not we need to punt a dio completion to a
workqueue on whether the IO is a write or not, add an explicit flag for
it. For now we treat them the same, reads always set the flags and async
writes do not.
No functional changes in this patch.
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
iocb->private is only used for polled IO, where the completer will
find the bio to poll through that field.
Assign it when we're submitting a polled bio, and get rid of the
dio->poll_bio indirection.
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Whether we have a write back cache and are using FUA or don't have
a write back cache at all is the same situation. Treat them the same.
This makes the IOMAP_DIO_WRITE_FUA name a bit misleading, as we have
two cases that provide stable writes:
1) Volatile write cache with FUA writes
2) Normal write without a volatile write cache
Rename that flag to IOMAP_DIO_STABLE_WRITE to make that clearer, and
update some of the FUA comments as well.
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
IOMAP_DIO_DIRTY shifts by 31 bits, which makes UBSAN unhappy. Clean up
all the defines by making the shifted value an unsigned value.
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reported-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Make the logic a bit easier to follow:
1) Add a release_bio out path, as everybody needs to touch that, and
have our bio ref check jump there if it's non-zero.
2) Add a kiocb local variable.
3) Add comments for each of the three conditions (sync, inline, or
async workqueue punt).
No functional changes in this patch.
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Convert platform_get_resource_byname() + devm_ioremap_resource() to a
single call to devm_platform_ioremap_resource_byname(), as this is
exactly what this function does.
Signed-off-by: Li Zetao <lizetao1@huawei.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
It is possible that dma_request_chan() returns EPROBE_DEFER, in which
case the driver defers probing without printing any message. Use
dev_err_probe() to record the probe deferral cause and ease debugging.
Signed-off-by: Wang Ming <machel@vivo.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
These options clearly turn *off* XSAVE YMM support. Correct the
typo.
Reported-by: Ben Hutchings <ben@decadent.org.uk>
Fixes: 553a5c03e9 ("x86/speculation: Add force option to GDS mitigation")
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Flushing the dirty buffer may take a long time if the cluster is
overloaded or if there is network issue. So we should ping the
MDSs periodically to keep alive, else the MDS will blocklist
the kclient.
Cc: stable@vger.kernel.org
Link: https://tracker.ceph.com/issues/61843
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Milind Changire <mchangir@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
